ABSTRACT
Background: Presently, the diagnosis of coronavirus-2019 (COVID-19) is a challenging task worldwide as the disease is spreading at a very faster rate when one person with the disease comes into contact with the other. Current information denotes that several people are detected with COVID-19 and the data analyst say that the rate of spread of the disease is increasing exponentially, across many countries in the world. Novelty: This investigation has facilitated the need for diagnosing the disease within a short duration of time by using the X-ray images of the lungs. This scheme deploys artificial intelligence like deep learning algorithms to diagnose COVID-19 among the affected people by maintaining social distancing. Real-time datasets are gathered from the government hospitals for those who are affected by COVID-19 and healthy people. Further investigation can direct the patients themselves to open the smart phone app which will record the respiratory sounds. Followed by this, the features are extracted using Discrete Wavelet Transform (DWT), where a threshold is applied to extract useful coefficients that can be used to train the deep learning neural networks using Fast Recurrent Convolutional Neural Networks (F-RCNN). The respiratory audio signals are captured to detect patients affected by coronavirus by a way of noncontact, nonintrusive approach. The results reported are valued in detection of COVID-19 by using a smart phone app which is available instantly. Objectives: This approach seems to be an indigenous, noninvasive, and cost-effective approach that will relive the patients from trauma of undergoing the swab test and awaiting the laboratory reports, which incurs time delay. Experimental results are obtained from 20,000 samples of patients suffering from COVID-19 and also persons who are normal. This mobile phone app is effective in diagnosing the COVID-19 from the X-ray images of the lungs. Even low-income people can also use this technology. Methods: The effectiveness of the proposed system which uses DWT, thresholding, and deep learning algorithms resulted with a performance whose F-measure is 96–98%. The classification is carried out to classify the COVID-19-positive and COVID-19-negative cases using Fast Recurrent Convolutional Neural Networks (F-RCNN). Expected Outcome: A smart phone app will be developed to detect the COVID-19 by using a noninvasive and easily affordable technique. The forecasted results were in the range of 89–95% for the above said algorithms. It is significant from the above results that the severe impact of COVID-19 can be diagnosed using a noninvasive mobile phone app using X-ray images. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.